Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; the at least two surrounding classes include a first surrounding class and a second surrounding class, a concentration of bone structures in proximity of the vessel in the first surrounding class is higher than a concentration of bone structures in proximity of the vessel in the second surrounding class; and if the surrounding of the vessel in the medical image is assigned to the first surrounding class, the segmentation of the vessel includes applying a learning-based ray casting algorithm to the medical image.
This invention relates to a method for segmenting and detecting the shapes of blood vessels in medical images, addressing challenges in accurately identifying vessels in complex anatomical regions with varying surrounding structures. The method involves classifying the area around a vessel in a medical image into at least two categories based on the concentration of nearby bone structures. The first category has a higher bone concentration, while the second has a lower concentration. Depending on the assigned category, the vessel is segmented using one of at least two predefined segmentation techniques. For regions with high bone concentration, a learning-based ray casting algorithm is applied to improve segmentation accuracy. The approach ensures that the segmentation method adapts to the surrounding anatomical context, enhancing precision in vessel detection. This adaptive technique is particularly useful in medical imaging where vessel visibility and structure can vary significantly due to adjacent tissues and bones. The method leverages machine learning and contextual analysis to optimize vessel segmentation in challenging imaging scenarios.
2. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; the at least two surrounding classes include a first surrounding class and a second surrounding class, a concentration of bone structures in proximity of the vessel in the first surrounding class is higher than a concentration of bone structures in proximity of the vessel in the second surrounding class; and if the surrounding of the vessel in the medical image is assigned to the second surrounding class, the step of segmenting includes applying an algorithm based on a morphological active contour without edges (MACWE) to the medical image.
This invention relates to a method for segmenting and detecting the shapes of blood vessels in medical images, addressing challenges in accurately identifying vessels in varying anatomical contexts. The method involves classifying the surrounding tissue of a vessel in a medical image using a first classifier, which categorizes the surrounding into at least two classes based on the concentration of bone structures near the vessel. The segmentation process then adapts based on the assigned class. For example, if the surrounding is classified as having a lower concentration of bone structures, the method applies a morphological active contour without edges (MACWE) algorithm for segmentation. Different segmentation techniques are predefined and selected based on the surrounding class to improve accuracy. The approach ensures that vessel segmentation accounts for anatomical variations, particularly differences in bone density, enhancing reliability in medical imaging analysis. The method is designed to optimize vessel detection in complex anatomical regions where traditional segmentation techniques may fail due to interference from surrounding structures.
3. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the first classifier is obtained by training a classifier with feature vectors for the at least two surrounding classes; and the first classifier is a k-nearest neighbor (KNN) classifier that assigns the surrounding of the vessel to a surrounding class most common among its k nearest neighbors, and k is a positive integer.
This invention relates to a method for segmenting and detecting the shapes of blood vessels in medical images, addressing challenges in accurately identifying vessel structures in varying anatomical contexts. The method involves classifying the surrounding tissue of a vessel in a medical image using a first classifier, which assigns the surrounding to one of at least two predefined classes. The vessel is then segmented based on the assigned surrounding class, with the segmentation method selected from at least two predefined approaches depending on the class. The first classifier is trained using feature vectors representing the surrounding classes and employs a k-nearest neighbor (KNN) algorithm, assigning the surrounding to the most common class among its k nearest neighbors, where k is a positive integer. The segmentation step adapts to the surrounding context, improving accuracy by tailoring the segmentation method to the specific anatomical environment. This approach enhances vessel detection in medical imaging by dynamically selecting optimal segmentation techniques based on local tissue characteristics.
4. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; the first classifier is obtained by training a classifier with feature vectors for the at least two surrounding classes; and the first classifier is a k-nearest neighbor (KNN) classifier that assigns the surrounding of the vessel to a surrounding class most common among its k nearest neighbors, and k is a positive integer.
This invention relates to a method for segmenting and detecting the shapes of blood vessels in medical images, addressing challenges in accurately identifying vessels in regions with varying anatomical structures, particularly where bone concentration affects visibility. The method involves classifying the surrounding area of a vessel in a medical image using a first classifier, which assigns the surrounding to one of at least two predefined classes based on the concentration of bone structures nearby. The classification step uses a k-nearest neighbor (KNN) classifier trained with feature vectors representing the surrounding classes. Once classified, the vessel is segmented using one of at least two predefined segmentation methods, selected based on the assigned surrounding class. This adaptive approach improves segmentation accuracy by tailoring the method to the specific anatomical context, such as distinguishing between regions with high or low bone density. The KNN classifier determines the surrounding class by identifying the most common class among its k nearest neighbors, where k is a positive integer. The method ensures robust vessel segmentation by dynamically adjusting the segmentation technique to the local anatomical environment.
5. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; the at least two surrounding classes include a first surrounding class and a second surrounding class, a concentration of bone structures in proximity of the vessel in the first surrounding class is higher than a concentration of bone structures in proximity of the vessel in the second surrounding class; the first classifier is obtained by training a classifier with feature vectors for the at least two surrounding classes; and the first classifier is a k-nearest neighbor (KNN) classifier that assigns the surrounding of the vessel to a surrounding class most common among its k nearest neighbors, and k is a positive integer.
This invention relates to a method for segmenting and detecting the shapes of blood vessels in medical images, particularly addressing challenges posed by varying anatomical surroundings, such as bone structures, which can interfere with accurate vessel segmentation. The method involves classifying the region around a vessel in a medical image into one of at least two surrounding classes based on the concentration of nearby bone structures. A first classifier, specifically a k-nearest neighbor (KNN) classifier, is trained using feature vectors representing these surrounding classes and assigns the vessel's surrounding to the most common class among its k nearest neighbors. Once classified, the vessel is segmented using one of at least two predefined segmentation methods, selected based on the assigned surrounding class. The two surrounding classes differ in bone structure concentration, with one class having higher bone proximity than the other. This approach ensures that segmentation adapts to the anatomical context, improving accuracy in regions with dense bone structures versus those with fewer bones. The method leverages the KNN classifier's ability to generalize from labeled training data, optimizing segmentation performance for different anatomical environments.
6. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; and the method further comprises the step of: classifying a shape of at least one section of a segmented vessel wall of the vessel by applying a second classifier to the segmented vessel wall, such that the at least one section of the segmented vessel wall is assigned to one of two shape classes.
This invention relates to the segmentation and shape detection of blood vessels in medical images, addressing challenges in accurately identifying vessel structures and their shapes in varying anatomical contexts. The method involves a two-step classification and segmentation process. First, the surrounding tissue of a vessel in a medical image is classified using a first classifier, assigning it to one of at least two predefined surrounding classes. The segmentation method for the vessel is then selected based on the assigned surrounding class, ensuring optimal segmentation for the specific anatomical context. The method employs at least two predefined segmentation techniques, each tailored to different surrounding tissue types. Additionally, the method includes a shape classification step where the segmented vessel wall is analyzed using a second classifier to determine the shape of at least one section of the vessel wall. This section is assigned to one of two shape classes, enabling detailed morphological analysis. The approach improves accuracy in vessel segmentation and shape detection by adapting to the surrounding tissue environment and refining shape classification post-segmentation. This method is particularly useful in medical imaging applications requiring precise vessel analysis, such as angiography or vascular disease diagnosis.
7. The method according to claim 6 , wherein the two shape classes include a first shape class relating to a vessel shape exhibiting a bifurcation.
This invention relates to a method for classifying shapes in medical imaging, particularly for identifying and analyzing vessel structures with bifurcations. The method addresses the challenge of accurately distinguishing vessel shapes in medical images, where bifurcations (branch points) complicate traditional shape analysis. The invention builds on a prior step of extracting shape features from medical images, such as vessel contours, and then categorizes these shapes into distinct classes. One of these classes specifically targets vessel shapes that exhibit bifurcations, allowing for specialized analysis or processing of these complex structures. The method may involve comparing extracted features against predefined criteria or models to determine whether a vessel shape belongs to the bifurcation-related class. This classification can improve diagnostic accuracy, treatment planning, or automated image analysis by ensuring bifurcated vessels are properly identified and handled. The approach is particularly useful in fields like angiography, where vessel morphology is critical for assessing cardiovascular conditions. By focusing on bifurcation detection, the method enhances the precision of medical imaging systems in analyzing vascular networks.
8. The method according to claim 7 , wherein the second classifier is obtained by training a classifier with one or more boundary descriptors for the two shape classes.
The invention relates to a method for classifying shapes, particularly in the context of computer vision or pattern recognition. The problem addressed is the accurate classification of shapes into distinct categories, which is challenging due to variations in shape boundaries and overlapping features between classes. The method involves using a second classifier that is specifically trained to distinguish between two shape classes. This classifier is trained using one or more boundary descriptors, which are features extracted from the edges or contours of the shapes. These boundary descriptors capture key characteristics of the shape boundaries that are relevant for distinguishing between the two classes. The second classifier leverages these descriptors to improve classification accuracy, particularly in cases where the shapes have similar or ambiguous features. The method may be part of a broader system where an initial classifier is used to perform a preliminary classification, and the second classifier refines the results by focusing on boundary descriptors. This two-stage approach enhances the robustness of the classification process, ensuring that shapes are correctly categorized even when their boundaries are complex or partially occluded. The use of boundary descriptors allows the classifier to focus on the most discriminative features, improving overall performance.
9. The method according to claim 8 , wherein the one or more boundary descriptors relate to at least one of an elongation, an eccentricity, a convexity, a solidity, and a standard deviation of radial distances of the segmented vessel wall.
This invention relates to medical imaging analysis, specifically for quantifying and characterizing blood vessel walls in medical images. The problem addressed is the need for accurate and automated assessment of vessel wall morphology, which is critical for diagnosing and monitoring vascular diseases. The invention provides a method for analyzing segmented vessel walls by extracting and evaluating boundary descriptors that quantify geometric and structural properties of the vessel walls. The method involves processing medical images to segment vessel walls and then analyzing the segmented boundaries to derive quantitative descriptors. These descriptors include elongation, which measures the length-to-width ratio of the vessel wall; eccentricity, which indicates how much the vessel wall deviates from a perfect circle; convexity, which assesses the degree of inward curvature; solidity, which measures the ratio of the area of the vessel wall to its convex hull; and the standard deviation of radial distances, which quantifies variations in wall thickness. These descriptors provide detailed insights into vessel wall morphology, enabling early detection of abnormalities such as aneurysms, atherosclerosis, or other vascular conditions. The method improves diagnostic accuracy by providing objective, reproducible metrics for vessel wall analysis, reducing reliance on subjective visual assessment.
10. The method according to claim 8 , wherein the second classifier is based on an approximate nearest neighbor (ANN) algorithm and a support vector machine (SVM) algorithm.
The invention relates to a method for classifying data using a hybrid approach combining approximate nearest neighbor (ANN) and support vector machine (SVM) algorithms. The method addresses the challenge of improving classification accuracy and efficiency in systems where traditional classifiers may struggle with high-dimensional or noisy data. The method involves a two-stage classification process. First, an initial classifier processes input data to generate preliminary classification results. This classifier may use any suitable technique, such as a decision tree, neural network, or other machine learning model, to filter or preprocess the data before further analysis. In the second stage, a hybrid classifier refines the preliminary results. This classifier combines an approximate nearest neighbor (ANN) algorithm with a support vector machine (SVM) algorithm. The ANN component efficiently identifies the most similar data points in a high-dimensional space, while the SVM component enhances decision boundaries and generalization. The hybrid approach leverages the strengths of both algorithms—ANN for speed and scalability, and SVM for robust classification—resulting in improved accuracy and performance. The method is particularly useful in applications requiring real-time processing, such as image recognition, fraud detection, or medical diagnostics, where both computational efficiency and high accuracy are critical. By integrating ANN and SVM, the method provides a balanced solution that addresses the limitations of using either algorithm alone.
11. The method according to claim 6 , wherein the second classifier is obtained by training a classifier with one or more boundary descriptors for the two shape classes.
This invention relates to a method for classifying shapes, particularly in scenarios where distinguishing between two shape classes is challenging due to overlapping or ambiguous features. The method addresses the problem of accurately classifying shapes when traditional classification techniques fail to reliably differentiate between the two classes. The solution involves using a second classifier that is specifically trained to handle boundary cases where the shapes are difficult to classify. The second classifier is trained using boundary descriptors, which are features or characteristics that lie near the decision boundary between the two shape classes. These boundary descriptors help the classifier focus on the most ambiguous or difficult-to-classify regions, improving its ability to distinguish between the two classes. The method leverages these boundary descriptors to refine the classification process, ensuring more accurate and reliable results. The first classifier, described in earlier parts of the invention, performs an initial classification of the shapes. However, when the first classifier encounters shapes that are difficult to classify, the second classifier is used to resolve the ambiguity. The second classifier is trained specifically to handle these boundary cases, ensuring that the overall classification system is robust and accurate. By combining the initial classification with the boundary-focused second classifier, the method improves the accuracy of shape classification, particularly in cases where the shapes are similar or lie near the decision boundary between the two classes. This approach is useful in applications such as medical imaging, object recognition, and quality control, where precise classification is critical.
12. The method according to claim 11 , wherein the one or more boundary descriptors relate to at least one of an elongation, an eccentricity, a convexity, a solidity, and a standard deviation of radial distances of the segmented vessel wall.
This invention relates to medical imaging analysis, specifically for extracting and characterizing vessel boundaries in medical images. The method addresses the challenge of accurately segmenting and quantifying vessel structures, such as blood vessels, in imaging data to support diagnostic and treatment decisions. The technique involves analyzing segmented vessel walls to derive boundary descriptors that provide quantitative insights into vessel morphology. The method processes medical images to segment vessel walls and then computes boundary descriptors that describe the geometric and structural properties of the segmented vessels. These descriptors include elongation, which measures the vessel's length-to-width ratio; eccentricity, which indicates how much the vessel deviates from a circular shape; convexity, which assesses the smoothness of the vessel boundary; solidity, which compares the area of the vessel to its convex hull; and the standard deviation of radial distances, which quantifies variations in vessel wall thickness. These descriptors help assess vessel health, detect abnormalities, and monitor disease progression. By quantifying these boundary characteristics, the method enables more precise and objective analysis of vessel structures, improving diagnostic accuracy and treatment planning. The descriptors can be used in various medical imaging modalities, including angiography, ultrasound, and MRI, to enhance the assessment of vascular conditions.
13. The method according to claim 11 , wherein the second classifier is based on an approximate nearest neighbor (ANN) algorithm and a support vector machine (SVM) algorithm.
This invention relates to a method for classifying data using a hybrid approach combining approximate nearest neighbor (ANN) and support vector machine (SVM) algorithms. The method addresses the challenge of improving classification accuracy and efficiency in systems where traditional classifiers may struggle with high-dimensional or noisy data. The method involves a two-stage classification process. First, an initial classifier processes input data to generate preliminary classification results. This classifier may use any suitable technique, such as a rule-based system or a simpler machine learning model, to filter or preprocess the data. The second stage applies a more sophisticated hybrid classifier that combines ANN and SVM algorithms. The ANN component efficiently narrows down potential matches by identifying the closest data points in a high-dimensional space, while the SVM component refines the classification by leveraging its strong generalization capabilities. This hybrid approach enhances accuracy by leveraging the strengths of both algorithms—ANN for speed and SVM for precision. The method is particularly useful in applications requiring real-time or near-real-time classification, such as image recognition, fraud detection, or medical diagnostics, where both computational efficiency and high accuracy are critical. By integrating ANN and SVM, the system achieves a balance between performance and reliability, making it suitable for complex classification tasks.
14. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; and the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; and the method further comprises the step of: classifying a shape of at least one section of a segmented vessel wall of the vessel by applying a second classifier to the segmented vessel wall, such that the at least one section of the segmented vessel wall is assigned to one of two shape classes.
This invention relates to the segmentation and shape detection of blood vessels in medical images, addressing challenges in accurately identifying vessel structures in varying anatomical contexts. The method involves classifying the surrounding tissue of a vessel in a medical image using a first classifier, which categorizes the surrounding into at least two classes based on the concentration of bone structures nearby. The vessel is then segmented using one of at least two predefined segmentation methods, selected based on the assigned surrounding class. This adaptive approach improves segmentation accuracy by tailoring the method to the local anatomical environment. Additionally, the method includes classifying the shape of at least one section of the segmented vessel wall using a second classifier, assigning the section to one of two shape classes. This step enhances the analysis by providing detailed morphological insights, which are critical for diagnosing vascular conditions. The invention combines contextual classification with adaptive segmentation and shape analysis to improve the reliability of blood vessel detection in medical imaging.
15. The method according to claim 14 , wherein the two shape classes include a first shape class relating to a vessel shape exhibiting a bifurcation.
This invention relates to a method for classifying shapes, particularly in medical imaging, to identify and analyze anatomical structures such as blood vessels. The method addresses the challenge of accurately distinguishing between different vessel shapes, including those with bifurcations, to improve diagnostic and treatment planning. The method involves processing image data to extract shape features of vessels and categorizing them into predefined shape classes. One of these classes specifically pertains to vessels exhibiting bifurcations, where the vessel splits into two or more branches. The classification is based on geometric and topological characteristics derived from the image data, enabling precise identification of bifurcated vessels. The method may also include additional steps such as segmenting the vessel structure from the image, extracting relevant features like curvature and branching points, and applying machine learning or statistical models to classify the vessel into the appropriate shape class. This classification aids in applications such as vascular disease detection, surgical planning, and monitoring of blood flow dynamics. By accurately identifying bifurcated vessels, the method enhances the reliability of medical imaging analysis, reducing errors in diagnosis and treatment. The approach is particularly useful in fields like cardiology, neurology, and radiology, where vessel morphology plays a critical role in patient care.
16. A method for segmentation and/or shape detection of blood vessels in medical images, the method comprising the steps of: classifying a surrounding of a vessel in a medical image by applying a first classifier to the medical image, the surrounding of the vessel being assigned to one of at least two surrounding classes; and segmenting the vessel depending on the surrounding class to which the surrounding of the vessel has been assigned; wherein a method used to perform the step of segmenting is selected from at least two predefined segmentation methods based on the surrounding class of the vessel assigned in the step of classifying; the at least two surrounding classes are different in a concentration of bone structures in proximity of the vessel; the at least two surrounding classes include a first surrounding class and a second surrounding class, a concentration of bone structures in proximity of the vessel in the first surrounding class is higher than a concentration of bone structures in proximity of the vessel in the second surrounding class; and the method further comprises the step of: classifying a shape of at least one section of a segmented vessel wall of the vessel by applying a second classifier to the segmented vessel wall, such that the at least one section of the segmented vessel wall is assigned to one of two shape classes.
This invention relates to a method for segmenting and detecting the shape of blood vessels in medical images, addressing challenges in accurately identifying vessels in regions with varying anatomical structures, particularly where bone concentration affects visibility. The method involves classifying the surrounding area of a vessel in a medical image using a first classifier, categorizing it into at least two classes based on bone structure concentration. The vessel is then segmented using one of at least two predefined segmentation methods, selected based on the surrounding class. This ensures optimal segmentation in areas with high bone concentration (first class) versus areas with lower bone concentration (second class). Additionally, the method includes classifying the shape of at least one section of the segmented vessel wall using a second classifier, assigning it to one of two shape classes. This approach improves vessel segmentation accuracy by adapting to local anatomical conditions and provides detailed shape analysis for diagnostic purposes. The technique is particularly useful in medical imaging where vessel visibility is obscured by surrounding bone structures.
17. The method according to claim 16 , wherein the two shape classes include a first shape class relating to a vessel shape exhibiting a bifurcation.
This invention relates to a method for classifying shapes in medical imaging, particularly for identifying and analyzing vessel structures such as blood vessels. The method addresses the challenge of accurately distinguishing between different vessel shapes in medical images, which is critical for diagnosing vascular conditions. The technique involves categorizing vessel shapes into predefined classes, with one class specifically designed to identify vessels exhibiting bifurcations—points where a single vessel splits into two or more branches. The method uses a machine learning model trained to recognize these bifurcation patterns, improving diagnostic accuracy by automating the detection of complex vessel structures. The classification process enhances the ability to analyze vascular networks, aiding in the detection of abnormalities like aneurysms or blockages. The method may also include preprocessing steps to enhance image quality and feature extraction to improve classification performance. By automating the identification of bifurcations, the technique reduces manual effort and increases consistency in medical imaging analysis. The invention is particularly useful in fields such as radiology and cardiovascular diagnostics, where precise vessel morphology assessment is essential.
Unknown
May 19, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.